Digital Twin (DT) technologies enable the creation of virtual replicas of learning environments, supporting personalized and real-time educational interventions. However, the integration of Artificial Intelligence (AI) within DT-enabled e-learning introduces critical challenges related to explainability, security, and learner privacy, and lacks a standardized operational framework. This study proposes and validates a comprehensive framework that operationalizes explainability and security for AI models in DT-based e-learning environments, balancing predictive performance, interpretability, and data protection. A quantitative experimental design involving approximately 300 learners evaluates three AI model variants: baseline, explainability-focused, and privacy/ security-enhanced. Predictive modeling employs temporal learner representations with ensemble predictors. Explainability is implemented through post-hoc interpretability techniques such as SHAP and Integrated Gradients, while privacy protection is ensured using Differential Privacy (DP) and Role-Based Access Control (RBAC). Multilevel mixed-effects models are utilized to assess predictive accuracy, explanation fidelity, and privacy guarantees, expressed as $\varepsilon $ -values. Results indicate that incorporating explainability mechanisms increases user trust by approximately 0.8–1.2 points on a Likert scale and enhances explanation fidelity by 25–30%. Integrating privacy controls produces a modest reduction in predictive AUC (up to 8%) but significantly mitigates data leakage risks. The proposed framework offers a standardized and reproducible evaluation suit for the certified deployment of explainable and secure AI systems in DT-enabled e-learning, facilitating transparent trade-offs between performance, interpretability, and privacy.
The integration of artificial intelligence (AI) and machine learning (ML) into education has transformed how student performance is predicted and monitored. Despite these advances, concerns regarding fairness, transparency, interpretability, and potential demographic bias remain significant challenges in educational prediction systems. This study developed ethically aligned and interpretable ML models for predicting student academic performance using only behavioural engagement and academic context variables. The open-access xAPI-Edu-Data dataset containing 480 student records was obtained from Kaggle. Four supervised algorithms, Multinomial Logistic Regression, Decision Tree, Random Forest, and XGBoost, were implemented in Python 3.11 using Scikit-learn, SHAP, and LIME frameworks. To minimise data leakage, all preprocessing operations, including standardisation and categorical encoding, were embedded within a Scikit-learn Pipeline fitted exclusively on the training folds. A stratified train/validation/test split (70%/15%/15%) with random_state = 42 was employed, while hyperparameters were optimised using five-fold cross-validation on the training and validation sets only. Model performance was evaluated using accuracy, macro F1-score, ROC-AUC, Brier Score, and Expected Calibration Error (ECE). Results showed that Random Forest achieved the best overall performance and calibration, while Logistic Regression provided superior interpretability. Across all models, visited_resources and raisedhands consistently emerged as the strongest predictors of academic achievement, emphasising the importance of student engagement behaviours. The study demonstrates that fairness-by-design prediction systems that exclude demographic variables can support equitable and actionable early-warning interventions. The novelty of the study lies in its ethically grounded and deployment-oriented framework for interpretable educational prediction in resource-constrained contexts.
M. A. Ayanwale, I. J. Chikezie· Journal of Computer Adaptive...· 0 citations
Trust-aware digital learning systems are essential for informed decision-making by educational stakeholders, as the growing complexity of digital learning environments increases the need for transparent and explainable learner profiles. While digital learning traces are crucial for identifying learner behaviors and categories, extracting meaningful profiles from large and complex data remains challenging due to bias, misinterpretation, and limited stakeholder guidance. Moreover, traditional learning analytics tools struggle to transform low-level traces into human-interpretable insights. In response to these shortcomings, we introduce a novel framework that exploits explainable artificial intelligence (XAI) to convert digital learning traces into meaningful and behavioral information. By integrating explainable machine learning techniques with human-centered design process, our framework connects fine-grained learning traces to high-level pedagogical interpretations. XAI-Profile is structured into three main components: (1) goal-driven requirements to decompose stakeholders’ goals into measurable subgoals linked to learner data, (2) visual analytics design to interpret learner profiles through transparent analytical artifacts, and (3) trust flow to support hypothesis testing and actionable, context-aware insights. Experimental evaluation in the écri+ project demonstrates that our framework generates interpretable learner profiles and validates pedagogical hypotheses. Educator-in-the-loop assessment confirms improved transparency and trust through interactive visual analytics. This work bridges XAI outputs with pedagogical insight, establishing a scalable, human-centered foundation for learning analytics.
Received: 9 April 2025 | Revised: 4 January 2026 | Accepted: 18 June 2026
Conflicts of Interest
The authors declare that they have no conflicts of interest related to this work.
Data Availability Statement
The data that support the findings of this study are openly available in GitHub at https://github.com/ouared14/XAI-Profile.
Author Contribution Statement
Abdelkader Ouared: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Madeth May: Conceptualization, Methodology, Validation, Investigation, Resources, Writing – review & editing, Visualization, Supervision, Project administration. Claudine Piau-Toffolon: Conceptualization, Methodology, Validation, Investigation, Resources, Writing – review & editing, Visualization, Supervision, Project administration. Nicolas Dugué: Conceptualization, Methodology, Validation, Investigation, Resources, Writing – review & editing, Visualization, Supervision.
Abdelkader Ouared, Madeth May, Claudine Piau-Toffolon et al.· Artificial Intelligence and...· 1 citation
This paper presents an end-to-end AI-driven personalised ideological education system integrating psychometrically-regularised knowledge tracing, hierarchical reinforcement learning, and mixed-initiative tutoring within a scalable cloud–edge deployment model. Using public proxy datasets comprising 2847 learner profiles and 1.24 million interaction records, we evaluated the proposed system’s knowledge-tracing, curriculum-sequencing, dialogue, affect-recognition, and system-performance components, the system achieved significant gains in learning efficiency and mastery (normalised learning gain +25.5% over expert-curated curricula) and high-fidelity student modelling (KT AUC = 0.821). Real-time responsiveness was sustained at scale, with latency-critical services remaining under 100 ms at p99 and end-to-end tutoring averaging 687.4 ms under 500 concurrent users. We further introduce governance, fairness, and non-persuasive oversight mechanisms to mitigate risks of value-shaping automation in civic and ideological contexts. These results provide component-level evidence that the proposed architecture can support adaptive sequencing under benchmark conditions. Real-world validation in ideological and civic education remains necessary before claims about classroom effectiveness or learner autonomy can be made.
Generative artificial intelligence (GenAI) is transforming the educational landscape by augmenting learning paradigms. However, state-of-the-art GenAI systems driving this transformation are predominantly developed and controlled by a small number of private companies; there is little clarity about their data retention practices and limited user control over inputs and outputs. In the context of education, end-users lack the awareness of how to safely adopt GenAI in learning. This raises significant concerns, particularly when proprietary or personally identifiable educational information may be shared with external GenAI platforms. In response to these concerns, universities are developing their own usage guidelines and policies to balance innovation with academic integrity, privacy, and security. Our research seeks to understand these emerging guidelines, with a particular focus on the privacy and security implications of integrating GenAI tools into academic environments -— an area that has received little attention to date. We conducted an in-depth qualitative analysis of GenAI-usage guidelines from 43 universities across 12 countries. Our findings reveal several key challenges, including barriers faced by universities in deploying privacy measures and adopting existing security frameworks. These insights lay the groundwork for designing more robust, privacy-aware GenAI guidelines for higher education.
Bei Yi Ng, Jiarui Li, X. Tong et al.· Proceedings on Privacy Enhan...· 6 citations
Background: The widespread adoption of Learning Management Systems (LMS) and digital learning platforms in higher education has produced vast repositories of student behavioural and academic data, creating unprecedented opportunities for predictive Learning Analytics (LA). However, academic failure, disengagement, and dropout remain persistent challenges that undermine institutional effectiveness and student wellbeing. Research problem: Existing Machine Learning (ML)-based student-risk prediction systems are predominantly optimised for predictive accuracy and operate as opaque "black boxes," offering limited insight into the reasons underlying a given risk classification. This lack of interpretability constrains their adoption by faculty, academic advisors, and administrators, who require transparent, actionable, and trustworthy evidence before intervening in a student's academic trajectory. Objective: This paper proposes and conceptually validates an Explainable Artificial Intelligence (XAI)-based Learning Analytics Framework designed to identify at-risk students at an early stage of a programme or semester while providing interpretable, human-understandable justifications for each prediction. Methodology: The framework integrates an eight-stage pipeline spanning data collection, preprocessing, feature engineering, comparative machine learning modelling (Logistic Regression, Random Forest, XGBoost, LightGBM, and a Multilayer Perceptron), rigorous evaluation emphasising recall and F1-score under class imbalance, post-hoc explainability using SHapley Additive exPlanations (SHAP) with a supplementary comparison to Local Interpretable Model-Agnostic Explanations (LIME), tiered risk classification, and structured educational intervention pathways. Contribution: The study synthesises 2020-2026 literature on learning analytics, educational data mining, dropout prediction, and explainable AI to identify a converging research gap concerning interpretability, fairness, and pedagogical actionability, and proposes a conceptual and methodological blueprint - including mathematical formulations, an algorithmic specification, and a proposed experimental protocol using publicly available higher-education datasets - that addresses this gap without prematurely claiming unverified empirical results. Educational significance: By coupling predictive analytics with transparent, instance-level and cohort-level explanations, the proposed framework is intended to support - rather than replace - the professional judgement of educators, enabling earlier, fairer, and more defensible academic interventions.
Ankit Kumar Singh, Rubi Singh, Mohd Nadeem· International Journal of Sci...· 0 citations
The advancements of modern artificial intelligence in education (AIEd) systems have greatly improved prediction accuracy and personal pacing. Traditional intelligent tutoring systems have the highest achievable aggregate predictive accuracy, a value that often is rooted in historical biases, mis-represents engagement signals from advantaged learner groups, and leaves vulnerable learner groups out of the picture. To address these challenges, we develop a mathematically robust framework of deep optimization with multiple objectives to ensure equity is maintained and carried forward across the tutoring lifecycle. We propose 5 main components: (i) demographic sensitivity gradient encoding (DSGE) for measures and limits direct demographic influence by computing gradient-level sensitivities of the learning loss with respect to latent demographic embeddings; (ii) counterfactual equity replay networks (CERN) which guided by the DSGE signal, CERN models the learning process through an explicit structural causal model and uses offline counterfactual simulation to quantify fairness-sensitive trajectory differences under stated identification assumptions. (iii) The engage-weighted fairness attention fusion dynamically balances student persistence and fairness risk, in order to not let high-level engagements mask concerns for fairness. (iv) Pareto-adaptive equity-accuracy-engagement optimizer adapts objective weights on the simplex space through a meta-gradient optimization which promotes stable convergence across the competing objectives under the adopted training procedure. In (v) equity-preserving policy distillation and validation, the high-capacity multi-objective model is compressed to a light-weight student model while ensuring that the student model preserves the equity of the multi-objective model when deployed. Our framework is validated using three real-world, publicly available educational datasets viz., EdNet, ASSISTments, and open university learning analytics dataset. Empirical results indicate that our framework achieves up to 67% lower learning gain disparity and 56% greater stability of dispersion in learning-gap distributions, while maintaining learning accuracy within 1.2% of the unconstrained, accuracy-only baselines.
A. Khan, Amit Pimpalkar, Tabassum H. Khan et al.· Scientific Reports· 0 citations
A new method, called CW-Net, translates the reasoning process of an autonomous vehicle’s AI system into understandable concepts that explain its behavior.